Business Challenge
In the production of single-use bioreactor bags, welding plays a critical role: sealing the polymer bag structure and integrating fluid ports, tubing, and connectors. These welds are typically made manually or semi-manually using heat-based welding machines.
Despite procedural controls, defects can occur due to:
Variability in machine performance
Operator fatigue or misalignment
Material inconsistencies in plastic components
These defects are often small, difficult to detect visually, and — if missed — can compromise downstream processes, contaminate product, or cause batch loss. A single bag may contain dozens of welds, requiring precise inspection without slowing down production.
For companies pursuing zero-defect strategies, this manual verification process presents both quality risk and operational inefficiency.
The Graypaw Solution: GrayVisions for Weld QA
Proconex engaged GRPAW to design a cognitive vision system capable of verifying weld quality immediately after fabrication.
Built on the GrayVisions platform, the solution integrates seamlessly into manual welding workstations. The operator simply places each welded bag or connector onto a designated inspection zone — a camera-equipped opening in the workbench — and GrayVisions handles the rest.
Core functionality:
Constant monitoring of the inspection zone via edge-connected camera
Automatic detection when a part is present and correctly positioned
Verification that the visual field is unobstructed (no hands, gloves, etc.)
Real-time inspection of weld zones using AI-based defect detection
Immediate visual feedback and classification on a tablet UI
The system uses controlled lighting and camera positioning to optimize defect visibility, ensuring consistent image quality regardless of ambient conditions.
Operator & Supervisor Experience
Operators interact with a streamlined QA interface, running on a nearby tablet:
- Real-time overlay of inspection results
- Highlighted defect areas, if present
- Accept/reject status with reasoning
- Option to flag or rescan parts before final logging
If a defect is found, the system automatically triggers alerts to supervisors, who can remotely review the flagged inspection — without entering the clean zone. Supervisors receive:
- Full video-linked inspection history
- Classification results and image evidence
- Decision tools for rework, rejection, or escalation
All results — good, bad, or reprocessed — are logged with full audit trail, including time, operator, device ID, and inspection outcome.
Outcome
Immediate post-weld inspection, reducing downstream risk
Reduced cognitive load and visual fatigue for welders
Improved first-pass quality and traceability
Supervisor visibility without entering sterile environments
Video-based audit trail of all inspections and decisions
“This is what real-time, in-process quality control should look like — zero disruption, total traceability.”
Blog
Our latest blog posts
Discover the latest insights, trends, and news from our expert team.
